declare @tTestTable1 table (
ID integer,
DayID date,
ArtID integer,
CntrID integer,
EndQnty decimal(16, 3)
)

Insert into @tTestTable1 values 
(1, '2018-07-29', 88, 8, 5.000),
(2, '2018-07-29', 55, 5, 5.000),
(3, '2018-07-30', 99, 9, 22.000),
(4, '2018-07-30', 88, 8, 20.000),
(5, '2018-07-30', 88, 8, 15.000),
(6, '2018-07-31', 88, 7, 10.000);

With Sales as
(Select *,case When lag(EndQnty) over (Partition by ArtID Order by DayID, ID) <= EndQnty Then 0 else 1 End as IsSale From @tTestTable1),
Groups as 
(Select *,sum(IsSale) Over (Partition by ArtID Order by DayID, ID) as Grp From Sales)

Select ID,DayID,ArtID,CntrID,EndQnty,
DATEDIFF(d,
isnull(min(DayID) over (Partition by ArtID,Grp Order by DayID, ID Rows between UNBOUNDED PRECEDING and 1 PRECEDING),--Ищем начало цепочки непродаж среди предыдущих значений
       lag(DayID) over (Partition by ArtID,IsSale Order by grp)),--Если не нашли, значит это и есть начало цепочки, поэтому берем дату предыдущей продажи
DayID) as DaysPrevSales
 From Groups
Order by ArtID,DayID,id